复杂环境下通用的手势识别方法研究计算机技术专业论文.docxVIP

ABST队CTAB ABST队CT AB STRACT Gesture recognition in complex background is difficult problem in human-computer interaction,and the existing method of gesture recognition has a large anaount of computation and long time consuming under the illumination change and background interference．To solve this problem，a general gesture recognition method is proposed,which carl improve the computational efficiency of gesture recognition．The whole method consists of three parts，gesture detection，gesture recognition and parameter updating Gesture recognition in complex background is a difficult problem in human-computer interaction，and the existing method of gesture recognition has large calculation and time consuming on the condition of influence of light and complex background．Aiming at the problem，a general gesture recognition method in complex environment Was proposed, which Can improve the computational efficiency of gesture recognition．The whole method consists of three parts，gesture detection，gesture recognition and parameter update In gesture detection section，in order to improve the computational efficiency a method that replace sliding window with bitwise operation in gesture detection Was proposed．The method realize binary support vector machine model and traverse the input map with shift operation supported by the 64 bit integer thus fulfill the task of rapid screening．The method does not need accurate extraction of the gesture area．It gets several possible windows in a very short time．The input of the gesture detection part is skin likelihood map．The image does not rely on the specific device，the general vision device is OK，the process of calculation needs 64 bit platform In gesture recognition section，a compute unified device architecture based convolution neural network was used to re·detect and identify the regions selected in the hand gesture detection．The target class of this model will include gesture classes and error classes because this step requires completi